The Use of GM(1,1) Model in Predicting Request Arrival Rate for Cloud-based Web Application

摘要

The primary purpose of this study was to use the GM(1,1) model to help the cloud-based web application provider to detect changes in the user access pattern. Accurate demand forecasting helps to make correct decisions for providing computing resources to meet dynamic demands by using cloud computing technology. A request rate forecasting approach based on modified GM(1,1) model is adopted in the Arrival Rate Predictor component of an adaptive resource provisioning mechanism. We present an optimal construction method of the background values of GM(1,1) model by minimizing Root Mean Square Error between the actual value and the predicted value. The optimal value of θ is dynamically determined by the current input data series to enhance the forecasting accuracy. Experimental results indicate that higher accuracy than the original model is achieved. In the Resource Demand Analyser component, a replica factor dynamic decision algorithm answering how many replicas would be provisioned is proposed. In contrast with the static resource provisioning policy based on statistical information of historical data, the acceptance rate of requests by using the proposed resource provisioning technology is consistently over 91%. These results demonstrate the forecasting approach based on the modified GM(1,1) model is an effective solution to predict the coming request rate.